--- title: "Thousands of AI idols debuted on Bilibili — and a rap by ancient Chinese poets went viral" date: 2026-10-19 category: Generative Media site: NeuroAI canonical: https://neuroai.site/a/na-aigc-ai-music-video-idols language: en --- # Thousands of AI idols debuted on Bilibili — and a rap by ancient Chinese poets went viral > Bilibili's "AI Idol Debut Season" drew nearly 10,000 AI music-video entries in months. A separate rap where historical figures vent about modern life racked up a million likes in a day — showing AIGC's next front is the music video, not just the song. A five-minute music video with a coherent story, believable lighting, and a hook you cannot stop humming — made by one person who typed **480,000 words of prompts** instead of hiring a crew. That is no longer a thought experiment on China's video platforms. It is a contest entry, and it is already mid-pack. ## A contest that turned prompts into pop stars In April 2026, Bilibili (B站) opened "AI Idol Debut Season" (AI偶像出道季), inviting creators to build virtual performers from scratch using generative tools. Within two months the platform had collected **nearly 10,000 submissions** and sparked **around 167,000 discussions**. One standout, a suspenseful narrative MV called *Pillow Man* (枕头人) by the creator EIKO (艾珂), was built from those 480,000 prompts and gathered close to a million views while drawing praise from seasoned music critics. The contest was not a fringe hobby. Established entertainment players showed up: a professional talent label co-organized the event, and the China Film AI Research Institute (中影人工智能研究院) and the Motion Picture Association (美国电影协会) joined as partners. The signal is that the music-video industry, not just bedroom creators, is taking AI idols seriously. ## The viral outlier: dead poets, living anxieties The clip that escaped the platform entirely arrived in October 2026. *The Experienced* (过来人), by creator 喆FILM, places ten historical and mythological figures — Du Fu, Su Shi, Li Bai, Cao Cao, the Monkey King — around a single microphone and lets them rap about modern life. Du Fu does not recite "my country in ruins"; he talks about rent and making a living. Su Shi does not boom "the great river flows east"; he talks about losing and still keeping a light on. The framing worked. Released across Weibo, Douyin, and WeChat Channels between October 13 and 15, it crossed **a million likes in a single day** and was quickly called the year's most viral AI song. The appeal was not the rendering quality — commenters noted the visuals were rough — but the writing. As one top comment put it: "sung straight into my heart." The creator's own rule, repeated in interviews, was "technology is the surface, the human heart is the core." ## Why the music video, not the song, is the breakpoint AI songwriting tools have existed for a while — Suno and Udio abroad, and Minimax (MiniMax), Mureka, and Kling (可灵) music models at home. Generating a pleasant track is no longer the hard part. The music *video* is where the bottleneck, and the opportunity, now sits: - **It fuses two hard problems.** A good AIMV (AI音乐视频) must sync visuals to a beat, keep a character recognizable, and tell a story in minutes — harder than either a still image or a raw audio clip. - **Tooling is catching up.** Dedicated AIMV platforms such as *Likemv* (立刻MV) shipped updates in 2026 that combine AI video generation with local video import, lowering the edit step. - **The weak link is consistency.** Creators report that beyond roughly three minutes, a single character's face can subtly drift between segments, forcing them to split long audio into shorter generation runs. In other words, the ceiling is no longer "can AI make a song?" but "can AI sustain a character and a narrative long enough to feel like a real artist?" ## Beyond the headline The Bilibili contest is best read alongside a broader cultural shift: the barrier to "making a music video" has collapsed from a budget to a prompt. That lowers the cost of entry dramatically, but it also floods the feed. When anyone can ship a polished AIMV, attention shifts from production value to concept — which is why the *Experienced* clip won on writing, not rendering. The same dynamic rewards writers who can direct and punishes those who expect the model to supply the soul. A global-reader caveat is translation. Much of this trend is unfolding on Chinese-language platforms with Chinese cultural references, and the humor of ancient poets rapping about rent does not always survive subtitles. Western audiences may meet AI idols through different frames — K-pop styling, gaming avatars, or fan-culture mascots — but the structural story is the same: the bottleneck is narrative consistency, not audio quality. Watch the storytellers, not just the spectacle. ## Trend reading According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, the rise of AIMV and AI idols shows creative power consolidating into the hands of writers who can direct, not just prompt — the winners will be storytellers who treat the model as a production crew, not a magic button. ## Honest limitations Several of the contest figures come from platform-coverage and self-media reporting rather than audited disclosures, so submission and view counts should be read as indicative, not exact. We deliberately excluded state-media political AIMVs from this analysis; they follow a different logic and were not part of the creator trend described here. We did not measure creator earnings, royalty treatment of AI-generated music, or training-data licensing — all unresolved questions. Character-consistency limits are drawn from creator anecdote, not a formal benchmark. ## What readers can do now - **Creators:** start with the writing. The viral hits won here because of concept and lyric, not render quality — draft the story before opening the generator. - **Music labels:** run a small AIMV pilot and watch retention by segment length; the three-minute consistency wall is where most projects quietly die. - **Viewers:** when an AI MV moves you, check whether it is the craft or the prompt — and credit the human director either way. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-aigc-ai-music-video-idols Free to quote with attribution and a link to the original.